Quality estimators aspire to quantify the perceptual resemblance but not the usefulness of a distorted image when compared to a reference natural image. However, humans can successfully accomplish tasks (e.g., object identification) using visibly distorted images that are not necessarily of high quality. This work investigates the usefulness (i.e., utility) of distorted natural images by 1) reporting methods to experimentally measure image perceived utility; 2) introducing and evaluating an objective utility estimator; and 3) generating useful but distorted natural images based on the proposed novel objective utility estimator. Subjective experiments were conducted to verify the distinction between the quality and utility of distorted natural images. Quality scores were obtained using a standard methodology. Novel experiments were conducted to collect responses from human observers regarding the usefulness of these distorted images, too. The resulting relationship between the utility and quality scores reveals that quality does not accurately predict utility. Distortions to high-frequency signal components of natural images are observed to have the greatest impact on utility. The experiment results demonstrate that a quality evaluation of a distorted image is different from its utility evaluation, so accurate quality estimators cannot accurately estimate utility. An understanding of the signal characteristics that distinguish utility from quality is obtained by analyzing and dismantling leading quality estimators, since no utility estimators exist. The natural image contour evaluation (NICE) is intro- duced as a utility estimator. NICE measures contour degradations of a distorted natural image relative to a reference natural image by extracting and comparing the edges from both images. Quality estimators and NICE are assessed as both quality and utility estimators. NICE provides accurate estimates of perceived utility scores and is argued to be compatible with shape-based theories of object perception. The perceived utility scores from the first set of experiments were found to exhibit limitations, and a novel technique that overcomes these limitations is proposed and implemented. The novel technique collects textual descriptions produced by observers viewing distorted natural images. The technique uses an observer-centric approach, so observers participating in the experiment dictate the relevant concepts that characterize image usefulness. This technique is used to obtain perceived utility scores for two collections of distorted images that simulate scenes captured by a surveillance system. The capability of both NICE and several leading quality estimators to estimate the perceived utility scores is reported. NICE is demonstrated to produce the most accurate estimates of perceived utility scores. Last, a procedure to generate useful distorted natural images based on NICE is presented. An image independent parametric quantization table that is compatible with baseline JPEG and based on NICE is provided. The quantization table is found by using a genetic algorithm heuristic search to perform rate-distortion optimization using a baseline JPEG encoder and NICE. Rate-distortion optimization using a genetic algorithm is discussed as a tool to analyze other objective estimators.