Image normalization of an image with large dynamic range

Question

I have an image of size 144*2209 and its dynamic range is large (from -1108 to 984).

I want to display this image and for that this range needs to be brought into 0 to 255 and for that I need to normalize the image.

Here lies the problem: when such large dynamic range is made compact, the values of pixels after normalization becomes very close to each other such that the image does not show up as it needs to be.

What can be done to resolve this problem???

Here is the link to the IMAGE.

Answer 1

You can use a linear transform to change the dynamic range of the original image, but be aware that you will be modifying the information of the image.

To do so, for a 8-bit range in Matlab, just use the following snippet :

bins = pow2(8); % = range
lin_eq_img = round( (bins - 1) * (img - min_img) / (max_img - min_img) );

But it will slightly affect the image :

Just a few remarks :

• even though your image dynamic range is above a 8-bit depth, it is not considered as 'a large dynamic range'
• depending on what and how you want to do, you might want to consider using the previous linear transform on a 16-bit dynamic in order to avoid losing details (by "squeezing" the pixels values distribution)
• you cannot want the intensity of your pixels values to be within the values given by 8-bit depth per pixel and say that you modified how your image is displayed : this is not a lossless operation !
• if you know what you want to enhance, there is plenty of non-linear transformations which could be helpful

Edit :

The 16-bit equalisation version will "keep your shades clearer" (no loss of details) but, of course, the rendered image will take more space. Here is a comparison :

I would strongly recommend you to perform the histogram normalization on a 2^16 intensity values span, in order to avoid details losses.