Sure. I assume by original you mean the raw data acquired by the sensor, and by 'thematic' you mean 'classification'. There were hundreds of them, back in the early 1970's when the only digital imagery was raw sensor data ( https://apps.dtic.mil/sti/pdfs/ADA040769.pdf ). This established the need for terrain, atmospheric column, sensor biases, illumination angle, cloud masking, and all the other pre- and post- processing needed to extract consistent useful information from scanline to scanline, scene to scene, and orbit to orbit, time of day, and seasonal fluctuations present in the real world. Otherwise, globally you end up with just a pile of noise.
Naser ahmadi sani Mr. Kader provided an excellent paper, see the section "TM SPECTRAL RESPONSE". But also consider the 'why' of what is described in the paper, the technological context of what and how information products were delivered. For instance, Adobe Photoshop even didn't exist yet at that time on desktops, and image manipulation and analysis still pretty much happened only at major agencies and firms. We literally have a million times more functional capacity on out laptops than was available at that time.