i've cleared my schedule for work on a new project, a watercolor of water waves, developed from photos shot during my recent europe vacation. this will be on an emperor sized (41" x 60") sheet of 600gsm Arches watercolor paper, and is the largest project i have yet attempted.
i will be posting photos and comments on the work as it progresses.
here's a thumbnail of the reference image. i won't go into the specific photo manipulation methods used to create it, as everyone has their own photoshopping procedures. in gist, the "curves" function was used to displace the photo value sequence, and various color adjustments were applied to develop the overall image appearance i was after. the file was saved in .gif mode (using the option "exact colors") to reduce the colors to a smaller number (14). finally the CIE Lab colors (shown in the color picker) were used to identify the color locations in a perceptual color space, and the pigment selection was developed from that information.
to begin the paint selection, i plotted the Lab location of the image colors in my artist's color wheel or CIELAB color map, in order to understand the basic color relationships and identify plausible pigment selections and paint mixtures. the colors grouped into muted blues, violets, reds and greens, many in complementary relationship, and in a repeated cycle through the spectrum.
in his watercolor tutorial (which i highly recommend), david dewey recommends the use of "color bars" to examine the color relationships within an image. my first color design step is usually to paint out the list of pigment selections as color bars, to get a first impression of how they interact. this also gives me an opportunity to adjust the colors, by glazing over the paints with other paints, and these changes are notated in the margin as mixture recipes.
although color bars fine for the broad impression, colors are often correlated or related within an image -- for example, if oranges cast violet shadows, then orange and violet are always contiguous in an image. this is hard to simulate with color bars. for that purpose, i used a grid of 3/4" squares, where the number of squares filled by a color approximately stands for its proportional area in the planned painting, and colors that go together in the painting can be painted next to each other in the grid.
the image below shows that the appearance of a paint selection is quite different when presented as color bars or as a color grid. (happily, the trex planks of my studio deck provide the perfect medium gray background for balanced color evaluation.)
as with the color bars, the color grid can be adjusted by glazing over certain colors, which suggests revised paint mixtures or pigment choices, which in turn can be used to create a new color grid ... and so on for as long as the color adjustments seem necessary. the color grid also lets me identify any problems that occur in edging the paints -- specifically one paint wicking or dissolving into another -- which affects both pigment choices and the sequence in which paints are applied.
the image below shows the initial color grid and a later version, and as you can see, the changes in the pigments made to eliminate granulation (a distraction in a complex image), whitening or blotching, and the adjustments in the paint mixtures, stand out nicely.
my procedure is to assign color locations freely in the first color grid, but in subsequent grids to copy this color allocation exactly, so that all visual changes in the pattern can be attributed solely to the colors, and not to the placement of the colors.
the final color grid can be used as standard color samples to match the mixture of paints, so that they can be applied at one pass to dry to the desired color. i want to avoid the chore of adjusting the paint colors after they have been applied!
next comes the underdrawing. in previous posts i've described my four preferred methods of developing the underdrawing for a painting: a freehand drawing of the subject, a freehand drawing corrected freehand (by photoshopping the image of the freehand drawing over a photograph of the subject); a rough charcoal outline drawn over a projected photo; and a graphite tracing of a projected photo.
all these methods seemed impractical given the size and complexity of the image. and printing out reference images, of 14 separate colors in 70 squares, would amount to 980 pages! so i've saved the reference image on a laptop computer, gridded to match the paper, with each color of paint as a separate image layer. this way i can paint the image freehand, color by color, square by square, using the laptop image as the reference.
by "blinking" each layer on and off, i could analyze how they fit together and identify the colors that follow the structural constraints or essential outlines in the image. these will be painted first. they define the crucial mapping from image to paper, and can be used as reference points to paint in the other colors.
finally, i measured the time it took me to copy the pattern for one color within one 6" area of the painting, then multiplied this by 70 (the number of six inch squares in an emperor sheet) and again by 14 (the number of different colors). this gave me an estimated time of 32 days, 5 hours a day, to complete the painting.
i decided this was doable.
i've ruled off the six inch squares on the sheet, chosen my paints and mixtures, and start work today.