0

i currently working on image enhancement of the endoscopic gastritis images. first step that i do is apply gaussian filter to the images. I had apply different size kernel. But i don't know how to measure is the enhanced image had been improved or not.

i had found some techniques like PSNR. But, it must has an original image. Now, i only have an endoscopic image and enhanced image by gaussian filter.

Can somebody tell me how can i measure the enhanced image is good or not?

Thank you.

user2545866
  • 31
  • 1
  • 4

1 Answers1

0

You may use SNR or CNR to measure the image quality. If the background signal in your image can be viewed as noise, calculate the standard deviation(delta) of those pixels. Then obtain the mean value of your contrast(u_gastritis), and SNR=u_gastritis/delta. If you have more than one contrasts in your image, CNR=(u_gastritis-u_2ndContrast)/delta You can also use Modulation transfer function to quantify the image quality, refer to: http://en.wikipedia.org/wiki/Modulation_transfer_function_(infrared_imaging)

lennon310
  • 12,503
  • 11
  • 43
  • 61
  • Hi, @lennon310 and thank you for your reply. currently, i use gaussian filter in the matlab to filter the image. The i find the difference between the no filtered image and filtered image by no filtered image minus filtered image. hence i get the only noise image. then, i calculate the standard deviation of the only noise image. Is it all i done true or not? But, i found that as the parameter of the gaussian filter increase, the standard deviation increase too. So, it cannot proved that the noise is reduced. what can i do now? please help!!! – user2545866 Nov 18 '13 at 06:29
  • 2
    @user2545866, i won't suggest you do the 'minus' operation on two images, since the filter/convolution process also change the gastritis values while suppressing the background noise. So the difference is not only the noise image. If the background region is prior knowledge, you can directly select those region to calculate the mean and std values. Hope it helps – lennon310 Nov 20 '13 at 19:41