Image Cropper JPEG re-encoding test


Loading an image into Image Cropper then saving it out again can potentially cause a loss of quality due to de-compressing and re-compressing, or re-encoding, the Jpeg Image.


This test was carried out by comparing Image Cropper to JpegCrops, a lossless Jpeg cropping program. Image Cropper is using the Lanczos resizing algorithm.

Magnified 240x160 pixel centre of the 1920x1080 cropped images


Left: Image Cropper                                         Right:  Jpegcrops lossless crop

The original image is 5184x3456 18MP jpeg best quality taken using a Canon 600D with a

24-105mm L series Lens. File size is 7.5MB  (uncompressed would be about 25MB).  


This image was then cropped to 1920x1080 with no re-sizing - a rather extreme crop but readily possible, the sort of thing that might be done for an HD video. The Image Cropper crop was saved as a jpeg at almost best quality (98) chosen to give the same file size as the lossless output file.


Note: Only certain ratios (multiples of 16 pixels) can be cropped losslessly and 1920x1080 Full HD is not one of them for this original size, so the nearest was taken.



      

  Original Image  5184x3456 pixels    7.5MB                    Magnified centre of the original image


                


  Cropped to 1920x1080 by Image Cropper                       Cropped to approx 1920x1080 by JpegCrops

  File size:    790KB                                                      File size:   798KB

  Quality :     98                                                          Quality:    as original (Canon ‘large’)

This clearly demonstrates that for a high quality jpeg image using Image Cropper to de-compress and re-compress causes no significant loss of quality when high quality settings are used. (85 up to 100 can be considered ‘high’). The limitation to such extreme cropping is the camera and it’s lens. This is not too surprising, its exactly what jpeg was designed to do.  


However, the situation changes significantly with low quality images.


The high quality original was saved using a jpeg quality of just 15 reducing the file size from 7.7MB to only 680KB. This was then cropped using Image Cropper and JpegCrops to 1920x1080.

 



Left:    Original LOW quality image                     Right:  Jpegcrops lossless 1920x1080 crop

                                                                                 File size: 60KB


Magnified 240x160 pixel centre. As you would expect the two are identical.




Left: Image Cropper saved at quality 85              Right:  Image Cropper saved at quality 15

        File size:   203KB                                                File size:  87KB



Clearly with a high degree of compression in the original image further re-compression is harmful but only if a lot of compression is used when saving. If saved with good Jpeg quality (85+) there is negligible extra loss. However it does result in a larger file size, even Q15 is larger.



Conclusion

For good quality Jpegs cropped and processed in Image Cropper there is little need to worry about quality loss if saved back with good quality.


If you need to work with highly compressed images AND retain a small file size a lossless cropper  may be a better choice, but if you just wish to crop and compress good images such as for a web page Image Cropper will do a good job.




Another very extreme test. The 1920x1080 jpeg image was loaded into Image Cropper and saved back to a jpeg at quality 98. The copy just saved was then loaded in and saved again. This was then done 10 times - load, save, load, save and so on.


This the the centre of the image after 10 generations, that is, 10 de-compress and re-compress processes:



Compare this with the first generation images above. The colour saturation has dropped very slightly but jpeg artifacts are no more apparent.

A quality of setting of 98 seems to maintain the original file size. At 99 it grew a little each time, 787KB to 899KB.  At 97 it shrank a little from 787KB to 748KB.


Although this may be only low compression it is still a big saving over a bitmap which would be about 6MB.

Download the 10th generation copy

Next Home Previous