LX5 versus LX3 High ISO RAW comparison shots here

Started Aug 30, 2010 | Discussions
brunobarolo
Contributing MemberPosts: 996
Like?
Re: Thanks David25...
In reply to David25, Aug 31, 2010

I have seen them, and they contribute to my conviction that I would use the LX5 in RAW mode only

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
New LX5 and LX3 at ISO 800 and 1600 - Part 1
In reply to David25, Aug 31, 2010

Well , IMO .......... the combination of optics and image-sensor on the new DMC-LX5 is (with little doubt) somewhat improved (enough to justify serious consideration of the DMC-LX5) !!!

The Panasonic-alleged 31% improvement of light-gathering ability and 38% improvement in "full-well-capacity" appear (not yet verified by actual tests to be performed by me today) to (both) have been committed for the (predicted) purpose of increasing the "efficiency" (apparent "brightness" to the eye) of the system - while (likely) retaining the "RAW Headroom" at the same (approximately) 1.0 EV value that the DMC-LX3 already has implemented.

The results of all this are impressive. While the display 100% crops represent (only) 5.00 Mpixels of the full 10 Mpixel sized full image-frame, the results are fairly impressive.

All of the same Silkypix settings (as previously listed in detail in my previous post displaying the original set of test-shots) were used. No re-sampling was performed. The 16-bit TIF output image-files from Silkypix were converted to a high quality (no chroma sub-sampling) 8-bit JPG using XnView 1.976.

LX3 - ISO=800 :

LX5 - ISO=800 :

LX3 - ISO=1600 :

LX5 - ISO=1600 :

The posted Lightroom 3.2 (I think it was) processing examples looked pretty nice when working from the previous set of RW2 source-images. It should be able to do an even nicer job with these newer, improved light-level and improved focus-integrity test-files.

I have seen Silkypix (SE and Pro) make a real noise/artifact mess out of the their 16-bit TIF output image files' 16-bit Logarithmic Histogram at relatively high histogram tone-levels when processing high ISO shots (as low as ISO=400.

However, I am impressed as to how smooth, clean, and (virtually) noise/artifact-free the 16-bit Log Histograms look when processing using the new release of Silkypix SE (Version 3.1.22). I was truly surprised to see how good the LX3 (as well the new LX5) ISO 800 and 1600 Silkypix output image-file histograms look.

I made a fair bit of "noise" (pun) about it on the Shortcut Software Silkypix Forum earlier in 2010 - including posting a fair number of very non-flattering 16-bit Log Histograms. Since, the forum has been completely dismantled ...

(I must say) that it appears that ISL may have actually "listened" about their Silkypix output noise/artifacts (especially when processing high ISO shots) ... Something positive is going on - and for that ISL deserves some credit ...

(Even) when subjected to Silkypix's characteristic "detail-softness" (likely a combination of the de-mosaicing algorithm used in the case of LX3 and LX5 RW2 image-files, as well as the system-internal (essentially "silent") Noise reduction that Silkypix has been said to likely perform in a number of locations in the processing-chain.

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
Re: New LX5 and LX3 at ISO 800 and 1600 - Part 2
In reply to Detail Man, Aug 31, 2010

Careful examination of the 16-bit Log Histograms for the test-shots above reveal that the LX3 provides (somewhat) more image-information in the lower tone-levels than does the LX5 (which "squeezes" the images "dynamic-range" more up towards the right-hand side of the histogram display). Thus, the LX3 images have a higher "contrast" (in the sense of having a wider tone-level range over which the image-information is distributed). This is not an insignificant fact (especially for "raw" shooters), as that wider tone-level "spread" of the LX3's image-data can be easily made the most of in a "raw" processing environment.

In a few previous posts I have noted of the better quality LX3/LX5 JPG comparison shots that the LX5 in-camera JPG "recipe" likely included a different Zero Exposure Level, as well as definitely employing a steeper "S-shaped" tone-curve in creating it's JPGs (as compared to the LX3 in-camera JPGs). This remains true, indeed.

In a manner not predicted, however, some amount of those (implicit, sensor-level) "tonal-characteristics" arise out of the new LX5 image-sensor, itself.

In the Log Histograms displayed below, one can get some feel for the "down-side" of using a steeper S-shaped tone-curve (in general). The image-data will have a reduced "image-dynamic-range" (because the S-curve essentially removes information at the bottom as well as the top tone-levels).

The histograms show the somewhat "narrower" distribution of tone-levels that the LX5 image-sensor system outputs (as compared to the LX3 image-sensor).

It is unknown (but possible, it seems) that there (may) be some sensor-level pre-"raw" processing going on in the LX5 image-sensor. I do not know. However, it is notable that the 1/1.7" Sony manufactured image-sensor in the Canon G11 as well as the Ricoh GRD III is thought to perform some "noise reduction" related sensor-level functions prior to the "raw" data level. So, such is not out of the question. In comparison with the previous LX3 image-sensor, there are no dramatic, glaring difference in image-detail - and (due I think to improved signal/noise ratio in the LX5 image-sensor), I would say that the LX5 image-sensor "wins-out" on fine-details when compared to the LX3 image-sensor.

It should be remembered that these test shots are both recorded at (what I consider to be) fairly high ISO Sensitivity levels. I would (in "raw" or JPG modes) personally not record above ISO=400 (but I am a bit of a "conservative" on creating image-noise problems in the very first-place).

It is possible that test-shots recorded at ISO=200 and ISO=400 might narrow the visible differences between the output if the LX5 and the LX3 image-sensors ...

However, overall (to my eyes, at ISO = 800 and ISO=1600), the LX5 output is improved over the LX3 in terms of visual "like-ability", having better and richer color-rendering" (at the "raw" data level), as well as a visibly improved signal/noise ratio which (to my eyes) translates to better definition of fine image-detail (a bit higher net spatial-frequency resolution, perhaps, via the optics combined with the image-sensor-level processing). The "raw" image-acuity is respectable and improved overall.

Note : The (normalized Exposure Level) version of this test-set yet to be posted is what will tell a more definitive story about the different "looks" of the RW2 images - when the Exposure Levels set by the two cameras are equalized ...
Stay tuned for that (today or tomorrow, most likely) ...

LX3 - ISO=800 16-bit Log Histogram

LX5 - ISO=800 16-bit Log Histogram

LX3 - ISO=1600 16-bit Log Histogram

LX5 - ISO=1600v 16-bit Log Histogram

Reply   Reply with quote   Complain
morepix
Veteran MemberPosts: 8,080Gear list
Like?
ISO 3200 or +1EV ??
In reply to David25, Aug 31, 2010

David25 wrote:

So, would you like ISO3200 or +1EV ?

I don't know what you mean by that. What I was trying to say is that one should not judge the ISO 1600 performance of the camera on the basis of an underexposed shot labeled ISO 1600.

And if I offended you by that, I should apologize. I can see you were trying to help.
--
http://www.pbase.com/morepix

 morepix's gear list:morepix's gear list
Panasonic Lumix DMC-LX7 Olympus Stylus 1 Olympus OM-D E-M10
Reply   Reply with quote   Complain
Swingline
Contributing MemberPosts: 690Gear list
Like?
Re: New LX5 and LX3 at ISO 800 and 1600 - Raw 6.2
In reply to Detail Man, Aug 31, 2010

I upgraded my CS5 raw to 6.2 which was released today. I see a little better detail and less noise at both iso's with the LX5 at 200%. At iso 800, I cannot read the text next to the barcode on the bottle with either camera but the LX5 looks a little better there as with the barcode. Both are cleanable at 1600 with Raw CS5 or Topaz Denoise5 to produce viewable pictures for non-pixel peepers.

 Swingline's gear list:Swingline's gear list
Panasonic Lumix DMC-LX3 Panasonic Lumix DMC-LX7 Olympus PEN E-PL5 Canon EF 35mm f/2.0 Canon EF 50mm f/1.8 II +11 more
Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
Addendum: New LX5 and LX3 at ISO 800 and 1600 - Part 1
In reply to Detail Man, Aug 31, 2010

Detail Man wrote:

(Even) when subjected to Silkypix's characteristic "detail-softness" [likely a combination of the de-mosaicing algorithm used in the case of LX3 and LX5 RW2 image-files, as well as the system-internal (essentially "silent") Noise reduction that Silkypix has been said to likely perform in a number of locations in the processing-chain] ...

... the LX5 as well as the LX3 test-shots look pretty good in terms of fine-detail, nevertheless ...

Results would be better when using Lightroom 3.x / Camera RAW 6.x.

Results would be stellar if DxO chooses to support DMC-LX5 RW2 and JPG image-files by creating a DxO Optical Corrections Module for the LX5 !!! ... At that point, the LX5 would be a "keeper" !!

Reply   Reply with quote   Complain
Ocean View
Regular MemberPosts: 287
Like?
Re: Addendum: New LX5 and LX3 at ISO 800 and 1600 - Part 1
In reply to Detail Man, Aug 31, 2010

Thanks for the analysis DetailMan.

I understand the desire for many to pixel peep. I think to some degree it helps them rationalize a new purchase. At least it does for me sometimes.

I recently saw a post (forgot posters name) where they took a sample of an LX5 shot and used DeNoise on it to remove noise.
I was so impressed with the results that I bought the plug in.

It made me realize one thing, noise at high ISO is not that important anymore as this plug in can do wonders.

I think regardless of which camera you have (LX3 or LX5) you will have to use some noise removal program on high ISO shots anyway so maybe all this pixel peeping is a moot point.

If I have the LX3, I don't think jumping to the LX5 is justified unless you need the extra reach and better video.

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
Normalized LX3 at ISO 800 and 1600
In reply to Detail Man, Aug 31, 2010

The composite average luminances of the LX3 and LX5 test-shots were measured using the Histogram tool in PSP 9.01. Silkypix "Exposure Bias" was adjusted to compensate :

ISO =800

LX3 = 88
LX5 = 100

LX5 Exposure Level is +0.184 Stops higher than LX3 Exposure Level

ISO =1600

LX3 = 90
LX5 = 104

LX5 Exposure Level is +0.209 Stops higher than LX3 Exposure Level

As a result of these measurement results, the LX3 (ISO=800 and ISO=1600) was re-processed in Silkypix using identical control-settings, and with the Exposure Bias control set to +0.2 EV in the case of both test-shots.

The pixel-size of the LX3 test-shots (and the specific cropped-image-frames) remain identical to the (non-normalized) previously posted LX3 test-shots.

(While I had judged the difference in exposure-levels as being higher than 0.2 Stops, the visually perceived "brightness" of the LX3 test-shots do appear to be close to the LX5 test-shots ... and the (newer, updated) 16-bit Log Histograms of the 16-bit TIF output image Files from Silkypix SE 3.1.22 now look virtually identical to those produced from the LX5.

My images can be downloaded at the "Original" (full pixel-size), so that interested readers can save the individual image-files to their systems for closer inspection using their JPG viewer of choice.

LX3 - ISO=800

LX3 - ISO=1600

LX3 - ISO=800 Log Histogram

LX3 - ISO=1600 Log Histogram

Maybe it's just that I own a DMC-LX3, but I would say that the (exposure-level-normalized) LX3 test-shots look a bit better normalized, and somewhat narrow the (for me) perceived differences between the LX5 and the LX3 samples.

One thing is (to my eyes) for sure. The DMC-LX3 RW2 + DxO Optics Pro 6.x solidly will outperform any of the in-camera JPGs that (either) the DMC-LX3 or the newer DMC-LX5 will ever be able to come close to producing.

Lightroom 3.x / Camera RAW 6.x presently support both the LX5 as well as the LX3.

However, (if) DxO decides to support DMC-LX5 RW2/JPG, there will be no equal to what will be able to be done with DMC-LX5 RW2 image-files.

I hope that this posted information will be useful to inform the readers/viewers ...

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
Re: Addendum: New LX5 and LX3 at ISO 800 and 1600 - Part 1
In reply to Ocean View, Aug 31, 2010

Ocean View wrote:
Thanks for the analysis DetailMan.

My pleasure. I've certainly been wondering myself about these differences.

I understand the desire for many to pixel peep. I think to some degree it helps them rationalize a new purchase. At least it does for me sometimes.

I recently saw a post (forgot posters name) where they took a sample of an LX5 shot and used DeNoise on it to remove noise.
I was so impressed with the results that I bought the plug in.

It made me realize one thing, noise at high ISO is not that important anymore as this plug in can do wonders.

I think regardless of which camera you have (LX3 or LX5) you will have to use some noise removal program on high ISO shots anyway so maybe all this pixel peeping is a moot point.

Some thoughts. For a high quality free plug-in, XiDenoiser features a number of different NR algorithms with (algorithm-specific) parameter controls. The Discrete Cosine Transform (DCT) algorithm is capable of very high-quality results. It is compatible with the (16-bit) PS, later PSPs, and Sagelight Image Editor, as well as the (8-bit) PSPs, GIMP, etc. It is a good-old ".8bf" file (that is not restricted to Adobe compatibility only). Just "drop it in" to the "Plugins" folder, and you are in business with a number of image-editors ... Easily found via Google. Safe, non-invasive, free ...

If one post-process JPGs as it stands, and have to farm-out the NR to a plug-in in the midst of the work-flow (hopefully early-on), it seems (to me) that at that level of complexity in the work-flow, one could "raw" process the camera's "raw" image-file , and (with DxO 6 or LR3 / CR6) absolutely transform the image-quality of the final (resulting) JPG ... Just thinkin' ...

If I have the LX3, I don't think jumping to the LX5 is justified unless you need the extra reach and better video.

I do not plan to purchase a DMC-LX5 until DxO (may) choose to support the camera. I have even considered (instead) purchasing a back-up DMC-LX3 to "cover my bases" (in case DxO does not support the DMC-LX5, and my currently owned LX3 should expire ...

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
Re: Normalized LX3 at ISO 800 and 1600
In reply to Detail Man, Sep 1, 2010

After doing more full-screen LX3 against LX5 comparison viewing (using the Exposure-Level normalized LX3 test-shots posted [so that the average (composite) luminance of the "raw" processed test-shots are essentially equal to the LX5 test-shots] ...

... I must say that the Signal/Noise ratio of the LX5 image-sensor is visibly better than that of the LX3 image-sensor .

Since the number of individual photo-sites (pixels), as well as the distance between them (pixel-pitch) remains unchanged in both image-sensors, I think that the perceived increases in the ability to resolve fine image-detal is a direct result of the improved Signal/Noise ratio of the new DMC-LX5 image-sensor.

I do get some sense of a slightly greater "smoothness" (of sorts) from the LX5 image-sensor that may well not arise out of it's improved Signal/Noise ratio. Perhaps this is an effect of the modified "micro-lens" assembly that exists over the surface of the individual image-sensor photo-sites.

It may be that the combination of the newer "micro-lens" assembly and/or any optical "anti-aliasing" (AA) filtering implemented on the LX5 image-sensor acts (in concert) to form a more effective (composite) anti-aliasing filtered-response .

This would attenuate spurious and unwanted image-artifacts resulting from spatial-frequency components that exceed the "Nyquist sampling criterion" limits - and might well act to "clean-up" the results a bit (while still potentially managing to provide somewhat better fine-details at spatial-frequencies that exist below the "Nyquist sampling limit").

This is a positive development. Aliased sampling artifacts do not improve the viewer's sens of "clarity" - they instead introduce "garbage" that does not accurately represent the optical image that is delivered to the image-sensor surface via the lens-system (on/in) the camera. In short - "more better" ...

I do not think that Panasonic has (in the jump from LX3 to LX5) done anything radically different with the LX5 image sensor (such as adding sensor-level pre-"raw" silent Noise Reduction functions, etc.). Hats off to Panasonic for that! ...

Regards, DM

Reply   Reply with quote   Complain
skiphunt13
Senior MemberPosts: 2,478
Like?
Re: LX5 versus LX3 High ISO RAW comparison shots here
In reply to David25, Sep 1, 2010

Er.. um... I've looked through all of these images a couple times. Is it me, or is the LX5 not really a significant jump up from the LX3 in overall image quality? A little better perhaps. And the longer lens is attractive.... but other than that, why upgrade?

-- hide signature --

Skip Hunt
Austin, Texas

http://skip-hunt.artistwebsites.com

Reply   Reply with quote   Complain
wakeskier
Regular MemberPosts: 368Gear list
Like?
Re: LX5 versus LX3 High ISO RAW comparison shots here
In reply to skiphunt13, Sep 1, 2010

skiphunt13 wrote:

Er.. um... I've looked through all of these images a couple times. Is it me, or is the LX5 not really a significant jump up from the LX3 in overall image quality? A little better perhaps. And the longer lens is attractive.... but other than that, why upgrade?

For me the upgrade in the video alone makes me want to upgrade.
The longer zoom is very nice too!

If the picture quality is as good as the LX3 (looks to be slightly better, at least at hifger ISO) I will be very happy.

If you don't care about video or extra zoom, then it's surely going to be harder to justify the upgrade.
It all depends on what is important to each individual

 wakeskier's gear list:wakeskier's gear list
Panasonic Lumix DMC-LX7 Panasonic Lumix DMC-LF1 Sony Alpha DSLR-A700 Sony 50mm F1.4 Tamron SP AF 17-50mm F/2.8 XR Di II LD Aspherical (IF) +5 more
Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
New LX5 at ISO+1600 with EC = +1.0 - Part 1
In reply to Detail Man, Sep 1, 2010

You know, what the LX5 needs is the DMC-FZ30/50 (35mm - 420mm) Leica lens-system, and an articulated LCD display-screen - because the new LX5 image sensor and LSI Chip-set (in "raw" data mode) is looking to be very nice, indeed. Call it the FZ-500, and a decent number of customers would line-up like restless cattle to purchase the camera (even if it cost around $600 USD or more) ...

Silkypix SE 3.122 not being my first "raw" processor choice (but the only on that I presently have running), I endeavored to ask the perennial pixel-peepitola question, ... "let's see, what happens when we push the noise-floor up by an amount nearly equal (but not greater than) the particular test-shot's actual 'RAW Headroom' when it is being 'raw' processed" ... just how bad does it get?

The implicit Signal/Noise ratio (Tonal-Range) of the image-sensor data remains fixed, but increasing the "Exposure Bias" so as to make maximal use of the 16-bit binary arithmetic range (minimizing the contributions of round-off-errors and other processing artifacts) makes optimal use of the "raw" processor (in terms of it's own processor Signal/Noise ratio as a contributing factor in the total resultant output Signal/Noise ratio of the image-file.

I found that the Silkypix Exposure Bias control could be set to +1.0 EV of positive exposure-compensation before it's (8-bit only) on-board (linear) Histogram display shows any appreciable "clipping" ("saturation") in a ny of the RGB color-channels.

Here are three easy to implement configurations of Silkypix:

White Balance and Contrast controls remain at "Camera Settings" (presumably manufacturer-state-secrets that Silkypix reads from the RW2 EXIF data).

Note: a Gamma correction constant = +1.15 is used in all cases when the "Camera Settings" option of the Contrast controls are selected.

The Color remains at the Standard setting (with Saturation at it's default value of 1.0).

False Color Control (essentially a chroma-specific NR-related function) remains at the default value = 30.

Demosaic Sharp remains at the default value =80.

"Exported" from Silkypix as a 16-bit TIF output image-file (100% crop). Not re-sampled. Converted to a high-quality (no chroma sub-sampling) JPG using XnView 1.976.

The differences in the control settings tried (displayed below):

"No Sharpness" (no Noise Reduction, no Sharpening) :

"Natural" mode (no Noise reduction, and relatively mild default Sharpening-control settings :

"Noise Reduction Priority [some Noise Reduction, and one-half as much "Outline Emphasis" as is used in the (also) relatively mild "Natural" mode control-settings] :

The 16-bit Logarithmic histograms (obtained using Guillermo Lujik's excellent and free Histogrammar 1.2 ) are all unremarkable (and look very good, free any significant sensor (or Silkypix) processing artifacts, with only a small number of "missing-levels" existing so far down in the 16-bit tonal range so as to be insignificant.

I must say that while the de-mosacing algorithm selected by Silkypix for use in processing DMC-LX5 RW2 image-files remains characteristically "soft" (clearly to my eyes out-performed by the de-mosaicing algorithms utilized by both DxO Optics Pro 6.x and Lightroom 3.x / Camera RAW 6.x), (also) causing high levels of chroma-noise (that can be mitigated fairly effectively using the Silkypix "False Color Control" function).

DxO (Chrominance and Luminance) NR is quite effective (and designed for) in high-ISO image-file processing - though the LR3/CR6 Chroma NR may be more effective on chroma-noise (a general feel from my DxO/LR3 processing experience).

The (so far not tried as I do not have the LR-3.2 update) LR3.x / CR6.x Chroma Noise control should (and, by the posted LR-3.2 processing looks to be) should be quite effective. If the LR-3.2 Sharpening controls are used at all (which sure would be nice, as th Sharpening tools do not come anywhere close to equaling what DxO "Lens Softness" optical corrections do so effectively for image-acuity), then (to my eyes) using some (Luminance) NR is very important (in order to reduce a (to me) quite ugly kind of "gritty", "quantized" noise/artifacts that become highly problematic where it comes to high-detail (high spatial-frequency) subject-matter (such as intricate foliage in the far-field in high Depth of Field nature/landscape shots) at 100% image-view. A genuine problem (for me) that requires use of Luminance NR (even when the image-file to be processed is a low ISO Sensitivity shot (such as ISO=80 through ISO=400) that (were it not for the LR 3.x processing noise/artifacts being produced) "should' (in my thinking) not require any Luminance NR ... End of LR/CR rant ...

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
Re: New LX5 at ISO+1600 with EC = +1.0 - Part 2
In reply to Detail Man, Sep 1, 2010

I do realize that all of the major "raw" processors (may well) apply some finite amount of NR (in some manner) - even when the user may imagine that NR functions are (by them) fully "diasbled". It may be that Adobe has (in changes in the Sharpening tool that are said by Eric Chan to have originated in the Lightroom 2.7 release) actually have placed all NR control in the user's hands. If so, they are to be commended for that. ( Perhaps ) the oft-required Luminance NR function in LR3/CR6 is simply doing the job that LR/CR (once) did, and that other "raw" processors are presently (silently) accomplishing without allowing a level of user NR control that enables complete disabling (or something very close).

Adobe's Eric Chan has stated that when the LR3/CR6 Sharpening "Detail" control-slider is set to Maximum (100), all sharpening effects are derived from Adobe's own "deconvolution-deblurring" function (perhaps still a "single-pass" deconvolution, as has been the case in the past, to the best of my knowledge derived from various sources who have published about these inner-workings).

Unfortunately, I have not been able to achieve good results using such a suggested configuration that can come anywhere close to the results provided by DxO Optics Pro 6.x "Lens Softness" corrections deconvolution-deblurring.

Life is so much easier in DxO. No array of (up to on the order of 6 interacting controls) to endlessly fiddle with in (perhaps, some) frustration. Simple:

(1) Adjust the "Lens Softness" controls (Global, Detail) to achieve the desired amount of "de-lens-blurring" and the corresponding enhancement of mage-acuity; then

(2) Make all of your other adjustments (Exposure, Low-Level Lighting, Color, etc.); then

(2) Adjust the "Noise Reduction" controls (Chrominance and Luminance) to taste.

Simple. Very low levels of control-interactivity. Do it once and you are usually done.

Reply   Reply with quote   Complain
Ocean View
Regular MemberPosts: 287
Like?
Re: LX5 versus LX3 High ISO RAW comparison shots here
In reply to skiphunt13, Sep 1, 2010

skiphunt13 wrote:

Er.. um... I've looked through all of these images a couple times. Is it me, or is the LX5 not really a significant jump up from the LX3 in overall image quality? A little better perhaps. And the longer lens is attractive.... but other than that, why upgrade?

-- hide signature --

Skip Hunt
Austin, Texas

http://skip-hunt.artistwebsites.com

If you compare just the quality of the images, then it's not worth the upgrade.

If you take everything into account and really think the new features are worth it, then it's worth the upgrade. Really a personal choice.

It's people like me that will have a hard time deciding cause I don't have either cams.
LX3 at $369 or LX5 at $500 decisions, decisions.

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,000
Like?
DMC-LX5 RW2 Custom Rendering
In reply to Detail Man, Sep 1, 2010

Not having Lightroom 3.2, I am presently "stuck with Silkypix SE 3.122" for a compatible "raw" processor. nevertheless, I thought that I would play with Silkpix and see what could be done for the DMC-LX5 ISO=1600 RW2 shot (the one with camera Exposure Compensation = 0.0, representing a worst-case Signal/Noise ratio compared to the other test-shots that were recorded witht the camera Exposure Compensation set to +0.33 EV).

Here is the result of what I was able to do in Silkypix (without using any Noise Reduction tools at all) :

Here is the result of what I was able (further) accomplish by applying the (free) XiMagic Denoiser 3.2 plug-in (DCT algorithm, Noise-Sigma=400, Cell-Size=4) running in the 16-bit processor/editor Sagelight 3.102. I then applied some mild USM in Sagelight (Pixel-Radius=0.5, Strength=100%), then converted to full 100% crop-size (not re-sampled) JPGs using XnView 1.976:

Not too bad for ISO=1600 on a 1/1.63" image sensor having 2.025 Micron pixel-pitch, and processed using (only) the free SE Version of Silkypix, the free XiMagic Denoiser, and the Sagelight 16-bit Image Editor version 3.102 (that was offered in April 2010 for a mere $1 USD promotional price; normally $40 USD cost) ...

In my nature/landscape application, I (still) would not dream of using an ISO Sensitivity as high as 1600. ISO=400 on the DMC-LX3 RW2image-files is just beginning to get into a Signal/Noise territory that is just beginning to approach "problematic" (though DxO NR can cope effectively with that noise without too much loss of image-detail).

It looks as if (when recording with the DMC-LX5 in RW2) one could shoot at ISO=400 with relative impunity, and (even) ISO=800 (might) work out in some less-critical cases (for those willing to engage in a bit more Noise Reduction and a bit less fine-detail integrity in processing) ...

Reply   Reply with quote   Complain
David25
Regular MemberPosts: 175Gear list
Like?
Landscape RAW shots for someone here
In reply to David25, Sep 1, 2010
Reply   Reply with quote   Complain
Vince1966
New MemberPosts: 1
Like?
Re: LX5 versus LX3 High ISO RAW comparison shots here
In reply to wakeskier, Sep 1, 2010

Same here, the quality and features in video mode are dramatically improved.
I got the LX3 because I found my DSLR to bulky on many occasions.
Interested in the LX5 for the same reason with my HDV camcorder.
Video mode on the LX3 is way to limited.

Reply   Reply with quote   Complain
Ketan G
Regular MemberPosts: 252
Like?
Re: Landscape RAW shots for someone here
In reply to David25, Sep 1, 2010

I took a stab at editing your landscape test shots. I've chosen the ISO 800 examples because this is as high as I would take a camera like this. I have made 100% crops in CS5. I opened the images in 16-bit Adobe RGB, made the edits listed below, then used the Save-For-Web feature to convert to a 100% quality sRGB JPEG.

These were converted in the newest version of ACR (6.2). My conversion settings were an attempt to match the color (more on that in a second), exposure (more coming), then recovery 10 and clarity +25, with all other settings left at default except that I turned off ALL noise reducation. Although I used recovery this example can't be used as a test of dynamic range because there were just too many differences between the files.

For color, I used the white balance dropper on the roof of the building in the crops below because the defaults were so different from each other. However, this set the LX5 at 5950K with -5 tint and the LX3 at 5100K and -9 tint. As for exposure, I tried to get the histograms close and eyeballed the image, which left the LX5 at +.1 and the LX3 at +.05. Looking at the results, the LX5 is a bit lighter in the shadows, which could be the result of DR in addition to my conversion; I probably could have upped its black level for similar contrast.

The original exposures were equal except for the shutter speed, 1/1250 for the LX5 and 1/1100 for the LX3, so all things being equal I think the LX5 gathers a bit more light in this example, not sure why.

The first set of 100% crops is are the LX3 and LX5 crops without any NR:

LX3:

LX5:

Then I treated both with the same amount of NR in Topaz DeNoise 5 then did a feature adjustment (contrast enhancement) in Topaz Detail 2. There is no USM.

LX3:

LX5:

My thoughts:

The untreated images have the LX3 showing more noise but also being a touch sharper than the LX5. In the samples with NR, I think both clean up well but the LX3 again is just a touch crisper in the end and I personally prefer its result. I don't think the differences in sharpness are 100% due to processing and camera NR because if you switch back and forth between the two you can see that the lenses are capturing the scene a bit differently as well. Specifically the LX3 seems to be stretched out a bit more, which may mean distortion is being corrected more in the LX5 sample. Either way, the extra sharpness in the LX3 shot doesn't result in a lot more fine detail and I doubt the difference would be apparent in anything but a very large print, if at all. Nonetheless it would have been nice if the LX5 was clearly better than the LX3 in this example.

I think waiting for an even more exacting test (tripod, shutter release, controlled lighting, etc) is the next step.

Reply   Reply with quote   Complain
Ketan G
Regular MemberPosts: 252
Like?
Re: Landscape RAW shots for someone here
In reply to Ketan G, Sep 1, 2010

Here's a second comparison. I did a conversion of the ISO 80 files but this time with all settings at default save turning off NR and a white balance using the dropper on the roof as before.

For this example, the original exposures had the same shutter speed, aperture, and of course ISO. However, the LX3 example was a bit darker than the LX5 example:

LX3:

LX5:

By adding 1/3 stop exposure compensation (+.33) to the LX3 image, its result was much closer to the LX5:

Again, the LX3 example is ever so slightly sharper, but its appearance is also a bit more coarse.

I then did a contrast enhancement of the LX5 crop with some CS5 Smart Sharpen (lens blur, strength 45. radius .2) and got this:

As you can see, although the LX3 shot starts sharper, a lot of that detail can be brought back with some editing on the LX5 file. This edit is a bit more intense than what I would normally do, but I overdid it a bit to make this point. Editing the LX3 file would also improve it, but I don't think the final results would be too far apart.

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads