Image quality assessment in retinal image compression systems

This paper presents a simple method to assess the quality of compressed retinal images. It is based on the idea that the retinal image should possess some common features which helps to define the quality of the image. Retinal blood vessels are important diagnostic features in ophthalmological images. Changes in retinal vessel structure helps in detecting the diabetic retinopathy(DR). The 2D discrete Laplacian operator is used to detect the vessel structure. The Laplacian error map shows spatial error distribution across an image. The overall image quality is given by Laplacian mean square error (LMSE). The experimental result shows that Laplacian error map localizes the error in a better way compared to structural similarity index (SSIM) map. Similarly, the objective error measure LMSE performs well in discriminating the images with different quality. The image quality is also evaluated using mean square error (MSE),peak signal to noise ratio (PSNR) and mean structural similarity index (MSSIM).