The detection of skin diseases is crucial due to their visibility and potential for transmission. However, clinical diagnosis presents challenges, often proving time-consuming and ineffective. Consequently, there is a pressing need for content-based image retrieval (CBIR) techniques to automate the detection of skin diseases. Effective multi-faceted ranking algorithms are widely used in CBIR, representing images through low-level features (color, texture, shape) and high-level features extracted from Convolutional Neural Networks (CNNs). Low-level features enable rapid detection of visual differences and are invariant to rotation and translation. Conversely, CNN-derived high-level features provide a richer semantic description. However, single query image systems may not fully capture image complexity, leading to inaccurate retrieval. This study introduces a novel method to improve the accuracy of skin disease image retrieval by combining ranking results from multiple query images. By integrating information from two input images, the proposed method enhances retrieval accuracy and efficiency. This approach leverages diverse semantic information for a more comprehensive and accurate retrieval process. Experiments on the Dermnet dataset demonstrate the effectiveness of this method in improving image retrieval quality.