IntelligentPhoto : A design for photo enhancement and human identification by histogram equalization, enhancing filters and Haar-Cascades

摘要

Flickr and Instagram are known to be an online crowdsourced photo sharing applications which people can share experiences with others. SmartPhoto is such a framework that was developed for this purpose. In the example of post-disaster recovery first responders survey the damage by taking pictures and then transfer them back to the rescue center. A problem with this system is how to abolish the redundancy and to find the most indicative photos containing human. SmartPhoto measure the utility of the crowdsourced photos based on the metadata of the photos which is accessible geometrical and geographical information. But this system did not address the problem with low quality images. If we enhance the low quality photos the perception factor of the photos get increased. The proposed system called IntelligentPhoto addresses this problem by first measuring the quality of the photos by checking the amount of blurriness and noisiness. It is based on by applying Laplacian kernel to the photos. The system then uses techniques known as Histogram Equalization to enhance the dark photos and perform other enhancing operations like noise reduction, smoothing, de-blurring, and edge enhancement to the low quality photos using wiener filter, median filter and smoothing filter. Selection algorithm are applied to find the most indicative photos and to abolish the redundancy in photos. Feature Feature matching is done to the selected photos by applying Haarcascades. Panoramas of the photos are constructed and it is send to rescue center for further action. This system is applicable to post-disaster recovery.