Image Size vs Quality

Most, if not all, digital cameras allow control over image sizeIn terms of megapixels and image qualityIn terms of compression. On some cameras, users chose image size and image quality independentlyExcept for lossless formats such as RAW and TIFF , while on others, a single option controls preset combinations of both. When both parameters can be controlled separately, people often wonder which combinations are more advantageous.

Obviously, the largest non-interpolated image size with the highest qualitylowest level of compression provides the best overall quality. To save space, one can either reduce image size, reduce image quality or both. Equally obvious is the fact that reducing both provides the least overall quality. The overall quality of reducing one parameter actually depends on the reduction step. When a 5 megapixel image is turned into a 4 megapixel one, there is a reduction of 20%. When it is turned into a 3 megapixel image, the reduction is 40%. While the reduction in size is quite clear when talking about image size, it is not as clear when talking about compression. Most cameras label the compression levels with subjective terms. The best indication of compression level should be found in the camera's manual or using a software such as JPEGQ.

What is important to know is the compression quality ratio for each image quality setting. What is frequently given instead is the compression storage ratio. There is a difference between the compression quality ratio and compression storage ratio. The former indicates relatively how much information is being discarded, the latter indicates how much storage is being saved. There is a difference between these two because most images have a certain amount of redundancy which can be eliminated without reducing quality. Therefore compressing image quality is worse than compressing storage by the same ratio.

Back to image size reduction. When reducing image size, the camera must apply a process called downsampling. Downsampling is the process of removing information from an image to produce an image with less pixels. Since information is lost during downsampling, it can be considered a form of compression. This form of compression is particularly bad because it is spatially uniform. This means that downsampling removes information across an entire image by the same amount.

For quality reduction the situation is usually different. Even if an image's quality is reduced by a certain factor, most lossy image compression technologies use perceptually based reduction. Perceptually based compression is better than downsampling because it strives to remove information less noticeable by the human visual system. A typical example of this is to compress color but not luminance because the human eye is more sensitive to contrast than hue.

The important point to underline here is that perceptually based compression is designed to be less visible than the same amount of non-perceptually based compression. Therefore, it is better to compress image quality than image size by the same amount.

Now for the simple math. Suppose a 5 megapixel camera can reduce its images to 3 megapixels. Suppose it also has two quality levels extra fine, which compresses quality by 9:1, and fine, which compresses quality by 12:1. Therefore, a fine 5 megapixel would be compressed 12:1. On the other hand, an extra-fine 3 megapixel image would be compressed in size by 5:3 and in quality by 9:1. The total compression of the extra-fine 3 megapixel image would therefore be 5:3 multiplied by 9:1 which is equal to 45:3 which is the same as 15:1. In this case, it is clear that the full-size image would have better overall quality. Unfortunately, the relative quality of image size and image compression combinations is not always clear when compression ratios are given relative to storage size. The reason for this is that a reduction in image size is usually the result of combining lossless compression and lossy perceptually-based compression. Therefore, the reduction in quality will be less than the corresponding reduction in size.

In conclusion, image quality compression is usually better than the same reduction percentage in image size. The important point to remember is that reduction in size, is a form of compression that is not optimized for our visual system and does not take advantage of redundancy within an image. When comparing overall image quality, the ideal is to calculate the total compression ratio. This measure will be more accurate than simply comparing image storage size or compression levels but, when all other things are equal, quality compression is better than downsampling.