Am I wrong on Panny 14-45?

alcelc wrote:

Thank you Eric, it sounds you have the answer.

Eric Nepean wrote:

The pedestrian is sharp, the text on the road is sharp, the building edges=s are sharp, the ridgeline is sharp.

Every bit of dark green foliage (in many parts of the image) is blurred.

Those were the mysterious to me.

This is not an optical problem. (or the all parts of the image would be affected)

This is not an image stabilisation problem. (or the all parts of the image would be affected)

This is an image processing problem.

But it is a SOOC jpg, as said had NR=0. As per my testing (under reasonable lighting condition) basically is close to NR=-5 but less noise. I keep the setting on all of my Pannys without problem as bad as this time.

I think the there is very little detail in the shadows because of the underexposure, and the JPEG conversion made it worse by reducing the dynamic range. On top of that there some noise reduction may be contibuting to lack of detail.

The underexposed green foliage makes up a big portion of the image, and so this is now very vissible.

Something likely had happened. If so, I might have also misunderstood the 12~32 under poor lighting condition.

Thank you for your head up.

I have found it very informative to read the description of JPEG compression in Wikipedia. Not the high level description - the example of the encoding process.

1. Convert from RGB to YCbCr (Y is luma, Cb and Cr are chroma)
2. Chroma subsampling - average the color values over 2x2 or 3x3 blocks of pixels (removes color detail)
3. Make the Discrete Cosine Transform of 16x16 blocks of pixels for each channel (result is a matrix)
4. Divide each element of the DCT result by each element of the Quantization matrix (the Quantization matrix suppresses "unnecessary" fine detail)
5. Round to the nearest integer (rounding removes detail, and reduces dynamic range, has the biggest effect in the shadows.)
6. Use run length encoding to losslessly compress what is left.

You have a large dark areas with uniform lightness (darkness) and with detail mostly in the color.

Step 2 removes half the color detail, Step 4 de-emphasizes fine detail, and then rounding in step 5 removes more detail and removes the bottom part of the dynamic range.

It is often said that the information discarded by JPEG compression is stuff the human eye doesn't see well; humans don't have good color perception in low light, we have black and white receptors for low light conditions.

My HDR merges always work significantly better when I start from raw images.