Surface emissivity maps in the infrared can contribute to explore Venus’ geology. Nightside radiance spectra at Themis Regio acquired by the IR mapping channel of the Visible and InfraRed Thermal Imaging Spectrometer (VIRTIS-M-IR) aboard Venus EXpress (VEX) are used to derive emissivity data from the three accessible spectral surface windows at 1.02, 1.10, and 1.18 μm. The measured spectra are simulated by applying a full radiative transfer model. Neglecting geologic activity, a multi-spectrum retrieval algorithm is utilized to determine the emissivity maps of the surface target as parameter vectors that are common to many spectrally resolved images that cover this target. Absolute emissivity values are difficult to obtain due to strong interferences from other parameters. The true emissivity mean of the target cannot be retrieved, nor can the emissivity mean of a retrieved map be strictly preset. The retrieved map can exhibit trends with latitude and topography that are probably artificial. Once the trends have been removed in a post-processing step, it can be observed that the magnitude of the resulting spatial emissivity fluctuations around their mean value increases with increasing mean value. A linear transformation is applied that converts the de-trended map to exhibit a defined emissivity mean value called reference emissivity, here 0.5, yielding the ‘renormalized emissivity map’ with accordingly transformed fluctuations. It is verified that renormalized emissivity maps are largely independent of the emissivity mean before renormalization, of modifications to interfering atmospheric, surface, and instrumental parameters, and of selected details of the retrieval pipeline and data calibration and preprocessing. Extremely large emissivity retrieval errors due to imperfect or unconsidered forward model parameters are effectively avoided. If the absolute emissivity at a given bin of the target were known, the absolute emissivity map of the entire target could be computed according to the mentioned transformation, assuming absent true trends with latitude and topography. Until then, the renormalized emissivities are interpreted as spatial variations relative to the reference emissivity. They represent an important step toward the retrieval of absolute emissivities. Renormalized emissivity maps of Themis Regio at the three surface windows are determined from 64 measurement repetitions. Retrieval errors are estimated by a statistical evaluation of maps derived from various disjoint selections of spectra and using different assumptions on the interfering parameters. Double standard deviation errors for the three surface windows amount to 3%, 8%, and 4%, respectively, allowing geologic interpretation. A comparison to results from an earlier error analysis based on synthetic spectra shows that unconsidered time variations of interfering atmospheric parameters are a major error source. Spatial variations of the 1.02 μm surface emissivity of 20% that correspond to the difference between unweathered granitic and basaltic rocks would be easily detectable, but such variations are ruled out for the studied target area. Emissivity anomalies of up to 8% are detected at both 1.02 and 1.18 μm. At present sensitivity, no anomalies are identified at 1.10 μm, but anomalies exceeding the determined error level can be excluded. With single standard deviation significance, all three maps show interesting spatial emissivity variations.