[Blog Post] MSI at last!

[Jorge_Stolfi]

Please excuse me for asking but have you downloaded one of the multispectral ones likeVoynich_71r+MB365UV_007_F.tif  because that one is very dark on my windows system

Those images were taken with a "technical" camera (which must capture accurate colors), not with a "consumer" camera (which must only look good, at any cost).  

Those images are 16-bits-per-pixel, but because of the light levels and/or the camera sensitivity the values in each image range from 0 to some "maxval" limit between 1000 and 4000 or so.  Thus, the brightest pixels are less than 4000/65535 = 6% of "white".  Which looks like black when viewed with any software designed for "consumer" images.  Yet those images have at least 10 times as many gray levels as  a common image.

I could not find the parameters of the images, but after a lot of mucking around I am guessing that the pixel values are proportional to physical brightness ("gamma = 1.0" in image processing parlance), and the You are not allowed to view links. Register or Login to view. images in particular have the following "white" pixel values>

Code:

# Spectral bands for page 116v1 - revised # Last edited on 2025-11-05 23:29:21 by stolfi # Line format: # #  "{band} {wlen} {illum}  {slo} {shi}  {wmd} {whi}  {vlo} {vhi} {maxval} # # See {band_table_funcs.py} for the meaning of these fields. # See ../00-Notebbok.txt for justifications for the {maxval} # band wlen illum  slo  shi  wmd whi  vlo vhi  maxval # {slo} = peak of histogram of scale patch 0. # {shi} = peak of histogram of scale patch 5. # # {wmd} = 0.99 quantile of grref patch. # {whi} = 0.99 quantile of whref patch. # {maxval} = {whi} MB365UV_007_F 365 0  65  265  1260 2120  55  725  2120 MB450RB_001_F 450 0  83 2225  1250 2373  113 1000  2373 MB470LB_002_F 470 0  83 2253  1263 2380  130 1097  2380 MB505CN_003_F 505 0  85 2337  1315 2465  160 1240  2465 MB535GN_004_F 535 0  85 2380  1365 2523  180 1360  2523 MB570AM_005_F 570 0  80 2250  1340 2425  225 1455  2425 MB625RD_006_F 625 0  80 2185  1305 2330  243 1485  2330 MB700IR_008_F 700 0  85 2040  1233 2180  360 1532  2180 MB735IR_009_F 735 0  93 2023  1233 2163  413 1560  2163 MB780IR_010_F 780 0  100 2055  1235 2190  475 1630  2190 MB870IR_011_F 870 0  110 2075  1297 2210  540 1687  2210 MB940IR_012_F 940 0  127 1993  1273 2125  567 1645  2125 MB365UV_029_F 365 1  20  167  855 1395  47  497  1395 MB450RB_023_F 450 1  67 1867  1115 2050  93  940  2050 MB470LB_024_F 470 1  70 1930  1145 2095  107 1045  2095 MB505CN_025_F 505 1  70 1920  1143 2087  123 1115  2087 MB535GN_026_F 535 1  73 1990  1195 2150  155 1225  2150 MB570AM_027_F 570 1  67 1875  1167 2063  215 1295  2063 MB625RD_028_F 625 1  67 1890  1183 2067  260 1355  2067 MB870IR_030_F 870 1  120 2300  1480 2453  755 1860  2453 MB365UV_037_F 365 2  65  180  855 1505  50  527  1505 MB450RB_031_F 450 2  63 1725  940 1865  90  857  1865 MB470LB_032_F 470 2  60 1690  915 1807  100  910  1807 MB505CN_033_F 505 2  67 1850  1010 1967  125 1087  1967 MB535GN_034_F 535 2  75 2020  1125 2165  160 1270  2165 MB570AM_035_F 570 2  67 1915  1105 2080  315 1355  2080 MB625RD_036_F 625 2  63 1717  985 1830  233 1267  1830 MB870IR_038_F 870 2  125 2300  1435 2507  705 2000  2507 # {whi} = 0.99 quantile of "txref.png" # # {maxval} = {whi} TX450RB_039_F 450 3    0    0    0 2667  150 1470  2667 # 3150  6621 TX535GN_040_F 535 3    0    0    0 1850  10  705  1850 # 2153  1211 TX570AM_041_F 570 3    0    0    0 1800  20  810  1800 # 2137  1194 TX625RD_042_F 625 3    0    0    0 1700  35  805  1700 # 2040  1020 TX700IR_043_F 700 3    0    0    0 2673  145 1495  2673 # 3325  1662 TX780IR_044_F 780 3    0    0    0 3333  215 1945  3333 # 4180  2090 TX870IR_045_F 870 3    0    0    0 2433  203 1485  2433 # 3085  1542 TX940IR_046_F 940 3    0    0    0 1595  170 1005  1595 # 2040  1020

Ignore anything after a "#".

The max effective pixel value (maxval) is the last column.  (As usual, this horrible bboard software messes up spaces even in code...)

The image I posted had its pixel values scaled from 0-2120 to 0-65535 and then to 0-255 for display.

The "MB" images are taken with light of a relatively narrow range of wavelengths centered on the number after the MB (in nanometers).  The three images with the same wavelength are taken with oblique light -- coming from the front and top of the page, from the front and bottom of the page, and with both lights at the same time.  From the shadows, it looks that the light is about 45 degrees from vertical, but with a large uncertainty.  The "TX" images are taken with transmitted light.  There are also "WB" images that may be combinations of various wavelengths, but I haven't looked at those.

Hope it helps, --stolfi