You are on page 1of 5

There is strong evidence that the layers in Obamas PDF image did not result from automatic (or

software) scanning. All of the objects have precisely placed rectangular boundaries, the sides of which touch on the boundaries of elements (i.e. text) of the objects. Both the vertical and the horizontal sides of the object boxes do this. Moreover for the vertical sides, the points of maximum extent of the text (or White speckles) occurs randomly on points along the vertical sides. For the horizontal sides in some cases the sides touch the bottom of a one or more pixels of text. No optimization software (that is based on automatic scanning of a graphical image) that I am aware of can "anticipate" where the points of maximum extent of each pixel of each text character will occur without first fixing the rectangular boundary of the selected region of text. The scanning (i.e. optimization) program would first have to determine the maximum extents of the pixel contents of each rectangular object in the horizontal and vertical directions before the rectangle object boundaries have been finalized. Then the program would adjust the left and bottom sides of each rectangular object to be congruent with the edges of these pixels. Thus the program would have to calculate the position of each side of each rectangle after the program has located the pixels which define the maximum extents of the contents. At a minimum, this would be a trial and error or iterative process. This is improbable. It is more likely that a human placed the object boundaries by a rectangle selection tool (or draw rectangle tool) in a graphics program. He would have used the "snap to point tools" to enable the precise placement of each rectangle. The observed object boundaries are more consistent with human manipulation than with automatic scanning. Heres the details on my earlier finding regarding the improbability that the nine objects with rectangular boundaries that comprise the Obama LFCOLB forgery were an artifact of file-size optimization. My earlier observation was that the edges of some of the object rectangles coincide with the maximim extent of the contained image. However this observation was not universally true for all nine objects and all four edges of each rectangular object. The details for each object are listed below. These results are from Adobe Illustrator CS6. img1 background + form matrix lines + sparse scattered text zero edges touch the maximum extent of the contained image. img2 form text + form check boxes + typewritten text + one partial signature + two handwritten dates left, right and bottom edges touch, top edge does not. img3 Onakas signature stamp left and bottom edges touch, top and right edges do not. img4 Onakas date stamp left and bottom edges touch, top and right edges do not. img5 Registrar Generals date stamp left and bottom edges touch, top and right edges do not.

img6 Local Registrars date stamp left and bottom edges touch, top and right edges do not. img7 "Non" left and bottom edges touch, top and right edges do not. img8 Top white splotches left, bottom and right edges touch, top edge does not. im9 Bottom white splotches left and bottom edges touch, top and right edges do not. The edges of the nine boxes which do not touch the maximum extents of the contained image vary in distance from the image extents. The background (img1) has a uniform white border around its four edges. The distances between the rectangular object boundaries, for the edges that do not touch pixels, and the nearest pixel of the contained image of each object are as follows for each of the nine rectangular object boundaries. The distances are in number of pixels for the particular object. The pixel size of the img1 are exactly twice the size of the pixels of the eight text or White speckles objects. The larger pixels are grayscale or green basket-weave image or white halo. The smaller pixels of the text objects are monochrome. Object img1 is the background layer. Img1 L 38 T 31 R 38 B 31 (cropped background image) img2 T 3 img3 T 2 R 3 img4 T 6 R 1 img5 T 5 R 7 img6 T 3 R 4 img7 T 5 R 3 img8 T 1 img9 T 5 R 6 (clipping path turned off) All of the other edges for each rectangle touch the edges of the pixels of the maximum extent characters of the contained images and are coincident with the edges of the pixels. For all nine objects, and for the object sides that do not touch the contained image, there is an integer number of pixels of white (or transparent) object background between the outermost pixels of the contained image and the object boundary. These results were obtained using the Adobe Illustrator CS6 vector graphics program and Photoshop CS6. These results indicate that a human set the object boundaries using the draw rectangle tool. The placement of the rectangular object boundaries was aided by the snap to grid tool working together with the snap to point tool. I am aware of no optimization software that could automatically set these same object boundaries. I have completed my analysis of the size and position of the nine object rectangles that comprise

the composite image of the WH LFCOLB PDF image. My efforts have focused on the central question Were the nine rectangles placed by man or machine? The overwhelming answer is that the placement was done by man. No file-size optimization software would have made the same placement decisions as are found in the WH LFCOLB PDF image. I had previously developed a method of re-creating the nine original images that resided on the forgers computer at the time that he created the WH LFCOLB PDF image using a "digital cut and paste" operation within his graphics program installed on his Macintosh computer. The nine recreated images are of larger size than their size within the WH LFCOLB PDF composite image. Additionally each rectangular image is rotated 90 degrees counter clockwise. Because the scale factors are different between the background and the non-background images, the composite image cannot be assembled using the nine re-created images in their original size. However, the scale factors were carefully chosen by the forger so that each re-created image has a pixel resolution of 72 PPI X 72 PPI. This is the same resolution as the screen resolution of the original Macintosh computer. However, the results of my analysis indicated that the forger was likely familiar with file optimization software and took advantage of his knowledge to make it look like the rectangle boundaries were set by machine. The results also indicate that the forger was familiar with modern digital graphics technology but was not familiar with 1961 vintage manual typewriters. Thus, most likely, the forger was relatively young. For my analysis, I chose an x,y coordinate system with origin at the upper left corner of the background layer. This point was therefore also the rulers origin. Consequently the x-axis is horizontal and x increases from left to right. The y-axis is vertical and y increases from top to bottom. This x,y coordinate convention is consistent with the most popular graphics programs that are currently provided by current sources. These include Adobe Photoshop CS6, Adobe Illustrator CS6, Adobe Acrobat XI Pro, and Inkscape. All of these programs set the rulers origin in the upper left corner of the screen and the (x,y) coordinates of each rectangle are measured from the origin to the upper-left corner of the rectangle. This approach assumes that the forger used a modern graphics program and a Macintosh computer to create his forgery. Results were obtained for two different pixel resolutions -- 300 PPI X 300 PPI and 150 PPI X 150 PPI. These were respectively the resolutions of the background layer and the eight nonbackground layers in the WH LFCOLB composite image. Three grids were analysed. These are 300 PPI X 300 PPI in blocks of 8 X 8, 300 PPI X 300 PPI in blocks of 16 X 16, and 150 PPI X 150 PPI in blocks of 8 X 8. Each of these three grids replicate for the re-created images at 72 PPI X 72 PPI resolution. Each of the nine rectangles satisfy the 8 MOD 0 condition for the top and right edges for the 300 PPI X 300 PPI resolution. Those edges of rectangles that satisfy the 16 MOD 0 condition for the 300 PPI X 300 PPI resolution also satisfy the 8 MOD 0 condition for the 150 PPI X 150 PPI. The modulo condition is satisfied whenever the quotient is an integer. Layers 2, 3, and 5 satisfy all three conditions for both the top and right edge of each rectangle.

Layer 6 satisfies all three conditions for the left and top edges. Oddly, the left edge of layer 6 satisfies all three modulo conditions. The exact positions of Layers 4, 8, and 9 are not critical. The "Non" in "None" layer satisfies the first condition but not the other two for the top and right edges. However, this object is constrained in both coordinate directions. Layers 1, 4, 7, 8 and 9 do not satisfy the second and third modulo condition for the right edge. However layer 1 (of these five layers) is constrained by the right edge of the page. Layers 4, 8, and 9 are not constrained in the x direction. And, as previously stated the "Non" layer is constrained in both x and y directions and therefore does not satisfy the second and third modulo condition for either the top or right edge. Finally the top edge of each of the nine layers satisfies all three conditions with the exception of the "Non" layer. Also the quotients are the same for the last two modulo conditions. The quotients for the last two conditions are exactly half of the respective quotients of the first condition. These results are consistent with the forger using a modern digital graphics program installed on a Macintosh computer to individually create each of the nine layers using a common screen resolution of 72 PPI X 72 PPI and 8 X 8 grid. The forger manually set each edge of each rectangular object boundary within an 8 X 8 screen grid at 72 PPI X 72 PPI resolution. He used the "snap to grid" tool in conjunction with the "draw rectangle" tool. All four edges of each rectangular object boundary were congruent with the grid lines. The top and right edges are always congruent with a major grid line (8 X 8 subdivisions). The left and bottom edges are congruent with a minor grid line. These two sides are also congruent with the edge of at least one pixel of at least one text character (or White speckle). These findings are conclusive that a human set the Width and Height and x,y position of each rectangular object boundary within the device space of his Macintosh Computer screen and using multiple grids that he created by means of a digital graphics program. The contrary view that a machine and/or file-size optimizer automatically set the size and position of each object rectangle within the PDF user space is indefensible. No combination of machine and optimization algorithm would need to apply multiple congruent grids to segment the scanned image. The results that follow are preliminary and are still being checked. If anyone finds a numerical error please post it and I will post a corrected page. See next page for my analysis results.

Resolution = 300 PPI X 300 PPI / 8 X 8 Blocks Top and Right Sides Obey 8 MOD 0 Left and Bottom Sides Align with Grid Lines (and also Touch Pixels) Layer N (x,y)(w,h)(x+w,y)((x+w)/8,y/8) 1 (0,0) (2552,3304) (2552,0) (319,0) 2 (373,880) (1819,1454) (2192,880) (274,110) 3 (1270,2848) (778,199) (2048,2848) (256,356) 4 (710,2928) (274,42)(984,2928) (123,366) 5 (1836,2160) (228,123) (2064,2160) (258,270) 6 (432,2240) (216,47) (648,2240) (81,280) 7 (1458,1960) (70,34) (1528,1960) (191,245) 8 (735,2528) (217,243) (952,2528) (119,316) 9 (1050,32) (142,132) (1192,32) (149,4) Resolution = 300 PPI X 300 PPI / 16 X 16 Blocks Layer N (x,y)(w,h)(x+w,y)((x+w)/16,y/16) 1 (0,0) (2552,3304) (2552,0) (159.5,0) 2 (373,880) (1819,1454) (2192,880) (137,55) 3(1270,2848) (778,199) (2048,2848) (128,178) 4 (710,2928) (274,42)(984,2928) (61.5,183) 5(1836,2160) (228,123) (2064,2160) (129,135) 6 (432,2240) (216,47) (648,2240) (40.5,140) 7(1458,1960) (70,34) (1528,1960) (95.5,122.5) 8 (735,2528) (217,243) (952,2528) (59.5,158) 9(1050,32) (142,132) (1192,32) (74.5,2) Resolution = 150 PPI X 150 PPI / 8 X 8 Blocks Layer N (x,y)(w,h)(x+w,y)((x+w)/8,y/8) 1 (0,0) (1276,1652) (1276,0) (159.5,0) 2 (186.5,440) (909.5,727) (1096,440) (137,55) 3 (635,1424) (389,99.5) (1024,1424) (128,178) 4 (355,1464) (137,21)(492,1464) (61.5,183) 5 (918,1080) (114,61.5) (1032,1080) (129,135) 6 (216,1120) (108,23.5) (324,1120) (40.5,140) 7 (729,980) (35,17) (764,980) (95.5,122.5) 8 (367.5,1264) (108.5,121.5) (476,1264) (59.5,158) 9 (525,16) (71,66) (596,16) (74.5,2) Background Mostly Text Onaka Signature Onaka Date Reg. Gen. Date Loc. Reg. Date Non Bottom Speckle Top Speckle Background Mostly Text Onaka Signature Reg. Gen. Date Loc. Reg. Date Non Bottom Speckle Top Speckle Background Mostly Text Onaka Signature Onaka Date Reg. Gen. Date Loc. Reg. Date Non Bottom Speckle Top Speckle

You might also like