The last decade brought computer vision to a more advanced state. It is the time when its research results started to influence virtually everybodyâs daily life, rather than being confined to industrial production lines or other niche applications. From user-interface applications (e.g. the Kinect), over face detection in photos and improved automated surveillance, to driver assistance, in all such areas computer vision contributed to the user-friendliness and safety of consumersâ environment. In this thesis we mainly focus on an important aspect of the algorithms that have brought those applications within reach, namely suitable data representations for computer vision. The main tasks that directly benefit are classification, detection, segmentation, and image enhancement. Performance, efficiency, and good trade-offs between those are the recurrent goals of our research. We consider the representation of vectorial features as linear combinations over pools of samples and this using sparse selections or ... all of them. More elaborate field representations are also considered for object class modeling and pixel labeling / image segmentation. The contributions include: 1. sparse representation-based projections for data dimensionality reduction. 2. sparse representations based on Iterative Nearest Neighbors, aimed at closing the gap between performance (sparse representations - the lasso type) and time efficiency (nearest neighbors). 3. Weighted Collaborative Representations based on the closed-form solutions of the Tikhonov regularization. 4. different image and feature level representations for Naive Bayes classification. 5. the Anchored Neighborhood Regression representation for fast example-based image super-resolution. 6. starting from sparse and collaborative representations, a training-free classification framework for textures, materials, and handwritten scores. 7. an Elastic Deformation Field Model for representing object classes for detection purposes. 8. by planar graph representation of images we use the Four Color Theorem from graph theory to lower the computational time of the loopy belief propagation in solving Markov Random Field labeling/segmentation problems in early vision. 9. efficient representation-driven contributions to applications such as traffic sign recognition and mapping, driver assistance, and extremely fast object detection.