Nowadays, information is not limited to textual representation but takes several other forms such as sketch-based image retrieval, where the user draws a query, and the system retrieves the most similar images. In this work we present a mixed approach combining shape context and Spark features, previously we had applied the Bag-of-Features strategy to select regions of interest, achieving significant improvement in effectiveness of the retrieval task. Our method works as a local strategy for key-points detection. Results are very auspicious, and we show different experiments conducted to demonstrate our proposed methodology. The highlight of this paper is the step-by-step description of the methodology to create a framework for sketch-based image retrieval.