I have noticed that IrfanView v4.57 x64 has an issue with reading very large JPEG-XR (.jxr) files.
It decodes them VERY slowly.
I conducted some test but I did not set aside a lot of my life for dealing with this issue.
I tested two editions of the same image.
The larger (original) image is 12656 x 14736 pixels, 24-bit RGB in BMP format.
The smaller (resized) image is 7186 x 8368 pixels, also 24-bit RGB in BMP format.
Both images were transcoded into PNG, JP2, WEBP, JLS and JXR formats (all lossless). All transcoding was done through IrfanView, except for JXR which was done through XnViewMP v0.98.1.
Each resulting image was then sequentially read by IrfanView and decoding times were recorded (by invoking "Image information" menu in IrfanView and reading the decoding time from there).
The RATIOS between decoding (reading) times between different image formats were THE SAME between all the mentioned compressed formats for BOTH the larger and the smaller image.
That means that the image resolution did not affect decoder's decoding speed (in pixels/s).
The only exception is JPEG XR.
With the smaller image, JPEG XR was about 20-25% faster to decode/read than JPEG-2000 (which is the slowest one) and about 9 times slower than PNG.
With the larger image, JPEG-XR became by far the slowest to decode, taking about 2,85x as much time as JPEG-2000 (the next slowest one) and about 19 times slower than PNG.
Conclusion:
It is obvious that there is either a megapixel value or an X or Y resolution value of an image above which JPEG-XR decoding performance suffers horrendeously for some reason.
I did not have the time to perform additional tests to determine the exact "breaking point" value, but I have at least determined that the problem exists.
P.S.: I have 32 GB of RAM and an SSD so fast that it almost disturbs the time-space continuum when I push it hard enough. Therefore, the hardware is not affecting the measurements.
It decodes them VERY slowly.
I conducted some test but I did not set aside a lot of my life for dealing with this issue.
I tested two editions of the same image.
The larger (original) image is 12656 x 14736 pixels, 24-bit RGB in BMP format.
The smaller (resized) image is 7186 x 8368 pixels, also 24-bit RGB in BMP format.
Both images were transcoded into PNG, JP2, WEBP, JLS and JXR formats (all lossless). All transcoding was done through IrfanView, except for JXR which was done through XnViewMP v0.98.1.
Each resulting image was then sequentially read by IrfanView and decoding times were recorded (by invoking "Image information" menu in IrfanView and reading the decoding time from there).
The RATIOS between decoding (reading) times between different image formats were THE SAME between all the mentioned compressed formats for BOTH the larger and the smaller image.
That means that the image resolution did not affect decoder's decoding speed (in pixels/s).
The only exception is JPEG XR.
With the smaller image, JPEG XR was about 20-25% faster to decode/read than JPEG-2000 (which is the slowest one) and about 9 times slower than PNG.
With the larger image, JPEG-XR became by far the slowest to decode, taking about 2,85x as much time as JPEG-2000 (the next slowest one) and about 19 times slower than PNG.
Conclusion:
It is obvious that there is either a megapixel value or an X or Y resolution value of an image above which JPEG-XR decoding performance suffers horrendeously for some reason.
I did not have the time to perform additional tests to determine the exact "breaking point" value, but I have at least determined that the problem exists.
P.S.: I have 32 GB of RAM and an SSD so fast that it almost disturbs the time-space continuum when I push it hard enough. Therefore, the hardware is not affecting the measurements.
Comment