The contrast sensitivity function (CSF) characterizes spatial detection in the human visual system and is typically measured from simple, synthetic stimuli. We used spatial frequency decomposition, RMS contrast modulation, a yes/no paradigm and an adaptive staircase to measure isolated and contextual CSFs (iCSFs and cCSFs) from natural images. We employed Barten’s mechanistic model and adapted it for contextual modeling purposes by postulating that, signal detection in a given frequency band, when presented amongst other broadband signals, can be modeled as if amongst noise. We found that the iCSF varies with pictorial content, but that the standard CSF model and the image’s contrast spectrums are sufficient to predict with relative success the cCSF for any given image. We finally discuss the suitability of cCSF models in image quality modeling.