Identifying Occluded Faces with a Modified Structural Similarity Index Measure

Facial obstructions, such as a pair of sunglasses masks, hats, and so on, have substantial implications when trying to reassemble the partly obscured regions of a face image. This study presents a new approach for removing occlusions based on the Structural Similarity Index Measure (SSIM). The side face is also the occluded images in a face image, reconstructing the occluded face is extremely difficult. The side face image problem is the main focus of this work. In order to find patterns in an obscured image, this study creates a Modified Structural Similarity Index Measure (MM-SSIM) that primarily focuses on side faces. This paper used two different techniques. First, the gallery dataset (X) and the masked dataset (Y) were each given their own dataset. Based on the occluded portion of the face image, a masked dataset was created. To deal with side face images, a Modified SSIM known as Mod-SSIM was created to strengthen the SSIM. Second, similar images from the gallery dataset were chosen as input images for Principal Component Analysis (PCA), reconstruction of obscured face images. The results of the experiments show that the modified SSIM, which was used with two datasets, FEI and Caspeal-R1, performs well and produces higher accuracy.