皮肤镜图像散焦模糊与光照不均混叠时的无参考质量评价

皮肤镜图像的采集质量直接影响后续的分析诊断结果.针对皮肤镜图像在采集过程中可能出现的散焦模糊和光 照不均两种类型的失真，提出了一种有效的无参考图像质量评价算法. 通过频率特性分析我们发现，散焦模糊主要 影响离散余弦变换（Discrete cosine transform，DCT）后的直流分量，而光照不均则主要影响第一交流分量.针对该特性，论文在频率域设计质量评价模型，首先将散焦模糊与光照不均两种失真成功分离，进而分别计算失真程度，给出评价指标.实验结 果表明，本文提出的算法既可以对散焦模糊或光照不均单一失真类型的皮肤镜图像进行质量评价，也可以对两种失真类型 同时存在的复杂皮肤图像进行评价，给出的评价结果稳定客观、互相独立，且与主观评价结果相一致.

Non-reference Quality Assessment of Dermoscopy Images with Defocus Blur and Uneven Illumination Distortion

The quality of dermoscopy images will impact the subsequent segmentation and analysis results. An efficient non-reference image quality assessment algorithm is proposed when defocus blur exists in an image with uneven illumination. The discrete cosine transform (DCT) is carried out and the frequency characteristic is deeply analyzed for dermoscopy images. It is found that defocus blur mainly affects the direct current (DC) component while uneven illumination mainly affects the first alternating current (AC) component. According to this finding, a quality assessment model is proposed in the frequency domain. Defocus blur signal is successfully separated from uneven illumination signal firstly, and then two metrics respectively for defocus blur and uneven illumination signals are calculated. Experiment results show that the proposed metrics can not only evaluate dermoscopy images with the single type of distortion caused by either defocus blur or uneven illumination, but also evaluate the ones with both of them at the same time. The evaluation results are stable, objective, independent, and consistent with the subjective evaluation results.