Automatic exposure estimation for exterior real estate photo editing in Adobe Photoshop

Manual exposure adjustment in Adobe Camera Raw (ACR) is a repetitive, time-consuming, and error-prone step in high-volume real estate photo editing. This paper presents a fully automatic exposure estimation method tailored to exterior real estate photography. We first identify two robust measurements: entropy and weighted variance of the luminance histogram. These metrics reliably characterise desirable tonal distribution under a fixed production ACR preset, and we empirically demonstrate that both are unimodal functions of exposure in this domain, ensuring a single well-defined optimum. This property is leveraged to develop an efficient directed search algorithm. The algorithm exploits the unimodality of the metrics and the practical observation that optimal corrections are typically small. As a result, it dramatically reduces the number of required Photoshop evaluations compared with exhaustive search while still locating the global optimum. The proposed method is completely unsupervised, requires no training data, and integrates seamlessly into existing ACR/Photoshop workflows. Experiments on large real production datasets show that the automatically estimated exposures are perceptually indistinguishable from those set by experienced human editors. They deliver the same visual quality while being both highly precise and substantially faster than manual editing, which eliminates one of the most labour-intensive steps in professional real estate post-production pipelines.