Enhancement of CSIFT Algorithm for Improved Feature Extraction in Psoriasis Image Analysis

The precise extraction of lesion features for psoriasis analysis is frequently hindered by the computational latency and "feature blindness" inherent in standard image processing algorithms, which often struggle to capture the subtle texture of organic skin lesions. Addressing the critical trade-off between execution efficiency and diagnostic sensitivity, an Enhanced Color Scale-Invariant Feature Transform (CSIFT) was engineered to optimize the feature extraction architecture for dermatological analysis through three pivotal technical interventions: the replacement of iterative pixel-wise computations with Vectorized Matrix Operations, the implementation of a Texture-Aware Detection module utilizing Adaptive Cr-Otsu Masking with Hybrid Harris Corner detection, and the integration of RootSIFT Normalization. Comparative performance analysis revealed a transformative improvement over the Standard method, as the proposed vectorization strategy achieved an 83.62% reduction in processing overhead, lowering the average execution time from 15,962.29 ms to 2,615.38 ms. Furthermore, the system resolved the loss of low-contrast details, recording a 218% increase in keypoint density from an average of 200.82 to 639.97 points while simultaneously improving the Matching Score to 97.06% compared to the Standard model’s 61.00%. These findings confirm that the Enhanced CSIFT provides a mathematically robust, accurate solution for dermatological imaging, effectively bridging the gap between computational speed and the precision required for clinical assessment.