Now a day images or pictures play a prominent role in the world. Content-based image retrieval (CBIR) systems are used to retrieve similar images based on query image. This system is capable of identifying similarities between query image and the set of images placed in the database. Even though it is an important research area from the last two decades, still there is scope for new technologies and algorithms to manipulate large amount of image databases in different fields like medical, social media and space images. Image contents are colors, texture and shape play significant role for image retrieval. All most all image retrieval systems are based on efficient feature extraction methods which are well-organized. This paper mainly concentrates to extract exact or most relevant images from the image database, here we compare query image with database image using pixels present in intensity vector. To accomplish this, the images are converted into gray scale. To achieve this, we are proposing a novel algorithm called as CEP (Count of Equivalent Pixel). This paper compares the existing sum of values of local histograms results with proposed system. The proposed system gives best results in retrieval of medical images in large database.