Enhancement of Veterinary Vaginoscopic Image of Female Canines using Color Transfer Algorithm in 1αβ Color Space

Abstract

Veterinarians need knowledge of the estrus stage of the reproductive cycle to decide the exact mating time for canine breeding. Through vaginoscopy, a visual examination of the vaginal vault of female canines can be done and the mating period can also be determined. However, due to low light illumination in vaginoscopic examination, the captured images lack visual quality. This hampers accurate determination of the mating period. In this study, an attempt has been made to enhance vaginoscopic images of female canines recorded at the College of Veterinary & Animal Sciences, Mannuthy, Kerala, India. These captured images are subjected to singular value decomposition (SVD), with the aid of a color transfer algorithm in 1αβ cone space. SVD results in a singular value matrix that accommodates the majority of the intensity information of the input image. By scaling these singular values, the total intensity of the input image is improved. Further, color transfer in 1αβ cone space is carried out to get better enhanced images. These enhanced images are evaluated using various quality parameters such as mean square error (MSE), peak signal to noise ratio (PSNR), entropy and mean. The proposed algorithm has an average MSE of 20.11, while PSNR, entropy and mean values are 35.65 dB, 7.16, and 104.15, respectively. The qualitative analysis shows that this algorithm can be used to enhance the veterinary vaginoscopic images intended for further processing. This algorithm facilitates veterinarians to clearly visualize the wrinkles and bloody discharge during oestrus cycle.