PROJECT TITLE :

Applications for no driven image quality assessment - 2016

ABSTRACT:

Goal: Dermoscopy pictures typically suffer from blur and uneven illumination distortions that occur during acquisition, that can adversely influence consequent automatic image analysis results on potential lesion objects. The aim of this paper is to deploy an algorithm that may automatically assess the standard of dermoscopy images. Such an algorithm may be used to direct image recapture or correction. Strategies: We describe an application-driven no-reference image quality assessment (IQA) model for dermoscopy pictures stricken by probably multiple distortions. For this purpose, we have a tendency to created a multiple distortion dataset of dermoscopy pictures impaired by varying degrees of blur and uneven illumination. The premise of this model is two single distortion IQA metrics that are sensitive to blur and uneven illumination, respectively. The outputs of those 2 metrics are combined to predict the standard of multiply distorted dermoscopy images using a fuzzy neural network. Not like traditional IQA algorithms, which use human subjective score as ground truth, here ground truth is driven by the application, and generated in step with the degree of influence of the distortions on lesion analysis. Results: The experimental results reveal that the proposed model delivers accurate and stable quality prediction results for dermoscopy pictures impaired by multiple distortions. Conclusion: The proposed model is effective for quality assessment of multiple distorted dermoscopy pictures. Significance: An application-driven concept for IQA is introduced, and at the identical time, a solution framework for the IQA of multiple distortions is proposed.